PropFLow

PropFlow — Retrofit intelligence for UK housing. Real EPC data, flood risk & planning constraints scored instantly by postcode. Discover where to act. For councils, landlords & agents.

Comment

Inspiration

Inspiration

The UK has a housing problem hiding in plain sight. 14 million homes sit below EPC Band C. A net zero target demands mass retrofit. And yet adoption is glacially slow.

We started asking why. The answer wasn't a lack of will or technology — it was a lack of actionable intelligence. The data exists: EPC ratings, flood risk, planning history, sale prices. But it's scattered across five government portals, incompatible in format, and impossible to act on without hiring a consultant.

We wanted to build the tool that makes that data speak. One postcode. One score. Instant clarity on where retrofit matters most — and why.

What it does

PropFlow is a three-layer retrofit intelligence platform:

  • 🔍 RetrofitIQ (built) — Enter any UK postcode and get an interactive map of real properties, each scored with a Retrofit Priority Score (RPS) built from live EPC ratings, flood risk zones, Land Registry sale prices, and planning constraints. Green = efficient. Red = act now.

  • 📊 RetroROI (designed) — Property-level analysis engine. Quantifies estimated value uplift, annual energy bill savings, CO₂ reduction, and planning approval probability for any specific home.

  • 📄 CarbonComply (designed) — Auto-generates a complete, blockchain-certified, planning-ready sustainability report. No consultant. No delay.

How we built it

PropFlow is a full-stack web application:

  • Backend: FastAPI (Python) REST API
  • Frontend: React.js with Leaflet.js for the interactive map
  • Database: MongoDB for flexible storage of EPC, planning, and price data
  • Data: Live calls to the MHCLG EPC Open Data API, HM Land Registry Price Paid dataset, and Environment Agency Flood Risk data

The Retrofit Priority Score is calculated as:

$$RPS = (0.35 \times E) + (0.25 \times F) + (0.20 \times P) + (0.20 \times H)$$

Where:

  • $E$ = EPC inefficiency concentration (proportion of E/F/G rated properties)
  • $F$ = Flood risk exposure (proportion in Flood Zone 2 or 3)
  • $P$ = Price paid delta (deviation below area median — proxy for underinvestment)
  • $H$ = Planning history friction (inverse of local authority approval rate)

RetroROI and CarbonComply are fully architecturally designed — RetroROI built around a machine learning model trained on Land Registry and EPC data, and CarbonComply using SHA-256 document hashing registered on an Ethereum-compatible blockchain for tamper-evident certification.

Challenges we ran into

  • EPC API inconsistency — Coverage gaps for non-residential and newly registered properties required a graceful mock fallback to keep the demo stable without surfacing false data

  • Scoring calibration — Getting the RPS weights to produce meaningful differentiation across diverse postcodes (inner-city new builds vs outer-London Victorian terraces) required real-world validation against known areas

  • 22-hour scope discipline — We made the deliberate call to build one layer exceptionally well rather than three layers poorly. That decision was harder than it sounds under time pressure

  • Data heterogeneity — Joining EPC, flood risk, price paid, and planning data across inconsistent postcode formats and update frequencies required non-trivial normalisation logic

Accomplishments that we're proud of

  • A fully working, live prototype pulling real data from three government APIs simultaneously

  • Meaningful score differentiation across postcodes — the RPS correctly distinguishes Canary Wharf new builds (~88) from Leytonstone Victorian terraces (~62) without any manual tuning

  • Zero broken demos — the mock fallback means PropFlow never crashes, even on non-residential postcodes like Buckingham Palace (SW1A 1AA)

  • Designing a complete, coherent three-layer product vision end-to-end within the hackathon window

  • Building something that addresses a genuinely underserved real-world problem — not a toy dataset, real open government data throughout

What we learned

  • Open government data is powerful but messy — The UK's open data ecosystem is genuinely rich, but format inconsistencies and API reliability gaps mean defensive engineering is non-negotiable

  • Weighted scoring is harder than it looks — Building a composite score that is mathematically sound and intuitively meaningful to non-technical users requires careful calibration and real-world sanity checks

  • Scope discipline wins — Shipping one working, demo-ready layer beats three half-built ones every time

  • The problem is real — Working through the use cases during the build made us realise how genuinely underserved councils and housing associations are for data-driven retrofit tooling

What's next for PropFlow

  • Train RetroROI on the full Land Registry and EPC dataset for ML-grade value uplift and savings predictions

  • Build the first CarbonComply PDF reports with blockchain certification

  • Live Article 4 and Conservation Area boundary API integration

  • National rollout across all 333 UK local authorities

  • Open API for housing associations, estate agents, and mortgage lenders

  • Green mortgage eligibility scoring integrated with lender platforms

Built With

Share this project:

Updates